Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/fs/open.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | */ | |
6 | ||
7 | #include <linux/string.h> | |
8 | #include <linux/mm.h> | |
1da177e4 | 9 | #include <linux/file.h> |
9f3acc31 | 10 | #include <linux/fdtable.h> |
0eeca283 | 11 | #include <linux/fsnotify.h> |
1da177e4 | 12 | #include <linux/module.h> |
1da177e4 LT |
13 | #include <linux/tty.h> |
14 | #include <linux/namei.h> | |
15 | #include <linux/backing-dev.h> | |
16f7e0fe | 16 | #include <linux/capability.h> |
086f7316 | 17 | #include <linux/securebits.h> |
1da177e4 LT |
18 | #include <linux/security.h> |
19 | #include <linux/mount.h> | |
5590ff0d | 20 | #include <linux/fcntl.h> |
5a0e3ad6 | 21 | #include <linux/slab.h> |
1da177e4 LT |
22 | #include <asm/uaccess.h> |
23 | #include <linux/fs.h> | |
ef3daeda | 24 | #include <linux/personality.h> |
1da177e4 LT |
25 | #include <linux/pagemap.h> |
26 | #include <linux/syscalls.h> | |
ab2af1f5 | 27 | #include <linux/rcupdate.h> |
73241ccc | 28 | #include <linux/audit.h> |
97ac7350 | 29 | #include <linux/falloc.h> |
5ad4e53b | 30 | #include <linux/fs_struct.h> |
b65a9cfc | 31 | #include <linux/ima.h> |
2dfc1cae | 32 | #include <linux/dnotify.h> |
3f6d078d | 33 | #include <linux/compat.h> |
1da177e4 | 34 | |
e81e3f4d EP |
35 | #include "internal.h" |
36 | ||
4a30131e N |
37 | int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs, |
38 | struct file *filp) | |
1da177e4 | 39 | { |
939a9421 | 40 | int ret; |
1da177e4 LT |
41 | struct iattr newattrs; |
42 | ||
43 | /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */ | |
44 | if (length < 0) | |
45 | return -EINVAL; | |
46 | ||
47 | newattrs.ia_size = length; | |
4a30131e | 48 | newattrs.ia_valid = ATTR_SIZE | time_attrs; |
cc4e69de MS |
49 | if (filp) { |
50 | newattrs.ia_file = filp; | |
51 | newattrs.ia_valid |= ATTR_FILE; | |
52 | } | |
1da177e4 | 53 | |
7b82dc0e | 54 | /* Remove suid/sgid on truncate too */ |
939a9421 AW |
55 | ret = should_remove_suid(dentry); |
56 | if (ret) | |
57 | newattrs.ia_valid |= ret | ATTR_FORCE; | |
7b82dc0e | 58 | |
1b1dcc1b | 59 | mutex_lock(&dentry->d_inode->i_mutex); |
27ac0ffe BF |
60 | /* Note any delegations or leases have already been broken: */ |
61 | ret = notify_change(dentry, &newattrs, NULL); | |
1b1dcc1b | 62 | mutex_unlock(&dentry->d_inode->i_mutex); |
939a9421 | 63 | return ret; |
1da177e4 LT |
64 | } |
65 | ||
a02de960 | 66 | long vfs_truncate(struct path *path, loff_t length) |
1da177e4 | 67 | { |
2d8f3038 | 68 | struct inode *inode; |
a02de960 | 69 | long error; |
1da177e4 | 70 | |
a02de960 | 71 | inode = path->dentry->d_inode; |
1da177e4 LT |
72 | |
73 | /* For directories it's -EISDIR, for other non-regulars - -EINVAL */ | |
1da177e4 | 74 | if (S_ISDIR(inode->i_mode)) |
a02de960 | 75 | return -EISDIR; |
1da177e4 | 76 | if (!S_ISREG(inode->i_mode)) |
a02de960 | 77 | return -EINVAL; |
1da177e4 | 78 | |
a02de960 | 79 | error = mnt_want_write(path->mnt); |
1da177e4 | 80 | if (error) |
a02de960 | 81 | goto out; |
1da177e4 | 82 | |
256984a8 | 83 | error = inode_permission(inode, MAY_WRITE); |
9ac9b847 DH |
84 | if (error) |
85 | goto mnt_drop_write_and_out; | |
1da177e4 LT |
86 | |
87 | error = -EPERM; | |
c82e42da | 88 | if (IS_APPEND(inode)) |
9ac9b847 | 89 | goto mnt_drop_write_and_out; |
1da177e4 | 90 | |
9700382c | 91 | error = get_write_access(inode); |
1da177e4 | 92 | if (error) |
9ac9b847 | 93 | goto mnt_drop_write_and_out; |
1da177e4 | 94 | |
9700382c | 95 | /* |
96 | * Make sure that there are no leases. get_write_access() protects | |
97 | * against the truncate racing with a lease-granting setlease(). | |
98 | */ | |
8737c930 | 99 | error = break_lease(inode, O_WRONLY); |
1da177e4 | 100 | if (error) |
9700382c | 101 | goto put_write_and_out; |
1da177e4 LT |
102 | |
103 | error = locks_verify_truncate(inode, NULL, length); | |
be6d3e56 | 104 | if (!error) |
a02de960 | 105 | error = security_path_truncate(path); |
907f4554 | 106 | if (!error) |
a02de960 | 107 | error = do_truncate(path->dentry, length, 0, NULL); |
1da177e4 | 108 | |
9700382c | 109 | put_write_and_out: |
110 | put_write_access(inode); | |
9ac9b847 | 111 | mnt_drop_write_and_out: |
a02de960 | 112 | mnt_drop_write(path->mnt); |
1da177e4 LT |
113 | out: |
114 | return error; | |
115 | } | |
a02de960 DH |
116 | EXPORT_SYMBOL_GPL(vfs_truncate); |
117 | ||
118 | static long do_sys_truncate(const char __user *pathname, loff_t length) | |
119 | { | |
48f7530d | 120 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
a02de960 DH |
121 | struct path path; |
122 | int error; | |
123 | ||
124 | if (length < 0) /* sorry, but loff_t says... */ | |
125 | return -EINVAL; | |
126 | ||
48f7530d JL |
127 | retry: |
128 | error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path); | |
a02de960 DH |
129 | if (!error) { |
130 | error = vfs_truncate(&path, length); | |
131 | path_put(&path); | |
132 | } | |
48f7530d JL |
133 | if (retry_estale(error, lookup_flags)) { |
134 | lookup_flags |= LOOKUP_REVAL; | |
135 | goto retry; | |
136 | } | |
a02de960 DH |
137 | return error; |
138 | } | |
1da177e4 | 139 | |
4fd8da8d | 140 | SYSCALL_DEFINE2(truncate, const char __user *, path, long, length) |
1da177e4 | 141 | { |
4fd8da8d | 142 | return do_sys_truncate(path, length); |
1da177e4 LT |
143 | } |
144 | ||
3f6d078d AV |
145 | #ifdef CONFIG_COMPAT |
146 | COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length) | |
147 | { | |
148 | return do_sys_truncate(path, length); | |
149 | } | |
150 | #endif | |
151 | ||
b01ec0ef | 152 | static long do_sys_ftruncate(unsigned int fd, loff_t length, int small) |
1da177e4 | 153 | { |
bf2965d5 | 154 | struct inode *inode; |
1da177e4 | 155 | struct dentry *dentry; |
2903ff01 | 156 | struct fd f; |
1da177e4 LT |
157 | int error; |
158 | ||
159 | error = -EINVAL; | |
160 | if (length < 0) | |
161 | goto out; | |
162 | error = -EBADF; | |
2903ff01 AV |
163 | f = fdget(fd); |
164 | if (!f.file) | |
1da177e4 LT |
165 | goto out; |
166 | ||
167 | /* explicitly opened as large or we are on 64-bit box */ | |
2903ff01 | 168 | if (f.file->f_flags & O_LARGEFILE) |
1da177e4 LT |
169 | small = 0; |
170 | ||
2903ff01 | 171 | dentry = f.file->f_path.dentry; |
1da177e4 LT |
172 | inode = dentry->d_inode; |
173 | error = -EINVAL; | |
2903ff01 | 174 | if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE)) |
1da177e4 LT |
175 | goto out_putf; |
176 | ||
177 | error = -EINVAL; | |
178 | /* Cannot ftruncate over 2^31 bytes without large file support */ | |
179 | if (small && length > MAX_NON_LFS) | |
180 | goto out_putf; | |
181 | ||
182 | error = -EPERM; | |
183 | if (IS_APPEND(inode)) | |
184 | goto out_putf; | |
185 | ||
14da9200 | 186 | sb_start_write(inode->i_sb); |
2903ff01 | 187 | error = locks_verify_truncate(inode, f.file, length); |
be6d3e56 | 188 | if (!error) |
2903ff01 | 189 | error = security_path_truncate(&f.file->f_path); |
1da177e4 | 190 | if (!error) |
2903ff01 | 191 | error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, f.file); |
14da9200 | 192 | sb_end_write(inode->i_sb); |
1da177e4 | 193 | out_putf: |
2903ff01 | 194 | fdput(f); |
1da177e4 LT |
195 | out: |
196 | return error; | |
197 | } | |
198 | ||
bdc480e3 | 199 | SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length) |
1da177e4 | 200 | { |
2cf09666 | 201 | return do_sys_ftruncate(fd, length, 1); |
1da177e4 LT |
202 | } |
203 | ||
3f6d078d AV |
204 | #ifdef CONFIG_COMPAT |
205 | COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_ulong_t, length) | |
206 | { | |
207 | return do_sys_ftruncate(fd, length, 1); | |
208 | } | |
209 | #endif | |
210 | ||
1da177e4 LT |
211 | /* LFS versions of truncate are only needed on 32 bit machines */ |
212 | #if BITS_PER_LONG == 32 | |
4a0fd5bf | 213 | SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length) |
1da177e4 LT |
214 | { |
215 | return do_sys_truncate(path, length); | |
216 | } | |
217 | ||
4a0fd5bf | 218 | SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length) |
1da177e4 | 219 | { |
2cf09666 | 220 | return do_sys_ftruncate(fd, length, 0); |
1da177e4 | 221 | } |
6673e0c3 | 222 | #endif /* BITS_PER_LONG == 32 */ |
1da177e4 | 223 | |
3e63cbb1 AJ |
224 | |
225 | int do_fallocate(struct file *file, int mode, loff_t offset, loff_t len) | |
97ac7350 | 226 | { |
496ad9aa | 227 | struct inode *inode = file_inode(file); |
3e63cbb1 | 228 | long ret; |
97ac7350 AA |
229 | |
230 | if (offset < 0 || len <= 0) | |
3e63cbb1 | 231 | return -EINVAL; |
97ac7350 AA |
232 | |
233 | /* Return error if mode is not supported */ | |
00f5e619 | 234 | if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | |
409332b6 LC |
235 | FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE)) |
236 | return -EOPNOTSUPP; | |
237 | ||
238 | /* Punch hole and zero range are mutually exclusive */ | |
239 | if ((mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) == | |
240 | (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) | |
79124f18 JB |
241 | return -EOPNOTSUPP; |
242 | ||
243 | /* Punch hole must have keep size set */ | |
244 | if ((mode & FALLOC_FL_PUNCH_HOLE) && | |
245 | !(mode & FALLOC_FL_KEEP_SIZE)) | |
3e63cbb1 | 246 | return -EOPNOTSUPP; |
97ac7350 | 247 | |
00f5e619 NJ |
248 | /* Collapse range should only be used exclusively. */ |
249 | if ((mode & FALLOC_FL_COLLAPSE_RANGE) && | |
250 | (mode & ~FALLOC_FL_COLLAPSE_RANGE)) | |
251 | return -EINVAL; | |
252 | ||
97ac7350 | 253 | if (!(file->f_mode & FMODE_WRITE)) |
3e63cbb1 | 254 | return -EBADF; |
1ca551c6 | 255 | |
00f5e619 | 256 | /* |
8fc61d92 | 257 | * We can only allow pure fallocate on append only files |
00f5e619 | 258 | */ |
8fc61d92 | 259 | if ((mode & ~FALLOC_FL_KEEP_SIZE) && IS_APPEND(inode)) |
1ca551c6 MS |
260 | return -EPERM; |
261 | ||
262 | if (IS_IMMUTABLE(inode)) | |
263 | return -EPERM; | |
264 | ||
0790b31b | 265 | /* |
6d2b6170 | 266 | * We cannot allow any fallocate operation on an active swapfile |
0790b31b LC |
267 | */ |
268 | if (IS_SWAPFILE(inode)) | |
6d2b6170 | 269 | return -ETXTBSY; |
0790b31b | 270 | |
97ac7350 AA |
271 | /* |
272 | * Revalidate the write permissions, in case security policy has | |
273 | * changed since the files were opened. | |
274 | */ | |
275 | ret = security_file_permission(file, MAY_WRITE); | |
276 | if (ret) | |
3e63cbb1 | 277 | return ret; |
97ac7350 | 278 | |
97ac7350 | 279 | if (S_ISFIFO(inode->i_mode)) |
3e63cbb1 | 280 | return -ESPIPE; |
97ac7350 | 281 | |
97ac7350 AA |
282 | /* |
283 | * Let individual file system decide if it supports preallocation | |
284 | * for directories or not. | |
285 | */ | |
286 | if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode)) | |
3e63cbb1 | 287 | return -ENODEV; |
97ac7350 | 288 | |
97ac7350 AA |
289 | /* Check for wrap through zero too */ |
290 | if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0)) | |
3e63cbb1 | 291 | return -EFBIG; |
97ac7350 | 292 | |
2fe17c10 | 293 | if (!file->f_op->fallocate) |
3e63cbb1 | 294 | return -EOPNOTSUPP; |
97ac7350 | 295 | |
14da9200 JK |
296 | sb_start_write(inode->i_sb); |
297 | ret = file->f_op->fallocate(file, mode, offset, len); | |
820c12d5 HS |
298 | |
299 | /* | |
300 | * Create inotify and fanotify events. | |
301 | * | |
302 | * To keep the logic simple always create events if fallocate succeeds. | |
303 | * This implies that events are even created if the file size remains | |
304 | * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE. | |
305 | */ | |
306 | if (ret == 0) | |
307 | fsnotify_modify(file); | |
308 | ||
14da9200 JK |
309 | sb_end_write(inode->i_sb); |
310 | return ret; | |
3e63cbb1 AJ |
311 | } |
312 | ||
4a0fd5bf | 313 | SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len) |
3e63cbb1 | 314 | { |
2903ff01 | 315 | struct fd f = fdget(fd); |
3e63cbb1 AJ |
316 | int error = -EBADF; |
317 | ||
2903ff01 AV |
318 | if (f.file) { |
319 | error = do_fallocate(f.file, mode, offset, len); | |
320 | fdput(f); | |
3e63cbb1 | 321 | } |
3e63cbb1 | 322 | return error; |
97ac7350 | 323 | } |
3e63cbb1 | 324 | |
1da177e4 LT |
325 | /* |
326 | * access() needs to use the real uid/gid, not the effective uid/gid. | |
327 | * We do this by temporarily clearing all FS-related capabilities and | |
328 | * switching the fsuid/fsgid around to the real ones. | |
329 | */ | |
6559eed8 | 330 | SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode) |
1da177e4 | 331 | { |
d84f4f99 DH |
332 | const struct cred *old_cred; |
333 | struct cred *override_cred; | |
2d8f3038 | 334 | struct path path; |
256984a8 | 335 | struct inode *inode; |
1da177e4 | 336 | int res; |
87fa5595 | 337 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
1da177e4 LT |
338 | |
339 | if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */ | |
340 | return -EINVAL; | |
341 | ||
d84f4f99 DH |
342 | override_cred = prepare_creds(); |
343 | if (!override_cred) | |
344 | return -ENOMEM; | |
1da177e4 | 345 | |
d84f4f99 DH |
346 | override_cred->fsuid = override_cred->uid; |
347 | override_cred->fsgid = override_cred->gid; | |
1da177e4 | 348 | |
086f7316 | 349 | if (!issecure(SECURE_NO_SETUID_FIXUP)) { |
1cdcbec1 | 350 | /* Clear the capabilities if we switch to a non-root user */ |
18815a18 EB |
351 | kuid_t root_uid = make_kuid(override_cred->user_ns, 0); |
352 | if (!uid_eq(override_cred->uid, root_uid)) | |
d84f4f99 | 353 | cap_clear(override_cred->cap_effective); |
086f7316 | 354 | else |
d84f4f99 DH |
355 | override_cred->cap_effective = |
356 | override_cred->cap_permitted; | |
086f7316 | 357 | } |
1da177e4 | 358 | |
d84f4f99 | 359 | old_cred = override_creds(override_cred); |
87fa5595 JL |
360 | retry: |
361 | res = user_path_at(dfd, filename, lookup_flags, &path); | |
6902d925 DH |
362 | if (res) |
363 | goto out; | |
364 | ||
2d8f3038 | 365 | inode = path.dentry->d_inode; |
256984a8 AV |
366 | |
367 | if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) { | |
30524472 AV |
368 | /* |
369 | * MAY_EXEC on regular files is denied if the fs is mounted | |
370 | * with the "noexec" flag. | |
371 | */ | |
372 | res = -EACCES; | |
2d8f3038 | 373 | if (path.mnt->mnt_flags & MNT_NOEXEC) |
30524472 AV |
374 | goto out_path_release; |
375 | } | |
376 | ||
256984a8 | 377 | res = inode_permission(inode, mode | MAY_ACCESS); |
6902d925 | 378 | /* SuS v2 requires we report a read only fs too */ |
256984a8 | 379 | if (res || !(mode & S_IWOTH) || special_file(inode->i_mode)) |
6902d925 | 380 | goto out_path_release; |
2f676cbc DH |
381 | /* |
382 | * This is a rare case where using __mnt_is_readonly() | |
383 | * is OK without a mnt_want/drop_write() pair. Since | |
384 | * no actual write to the fs is performed here, we do | |
385 | * not need to telegraph to that to anyone. | |
386 | * | |
387 | * By doing this, we accept that this access is | |
388 | * inherently racy and know that the fs may change | |
389 | * state before we even see this result. | |
390 | */ | |
2d8f3038 | 391 | if (__mnt_is_readonly(path.mnt)) |
6902d925 | 392 | res = -EROFS; |
1da177e4 | 393 | |
6902d925 | 394 | out_path_release: |
2d8f3038 | 395 | path_put(&path); |
87fa5595 JL |
396 | if (retry_estale(res, lookup_flags)) { |
397 | lookup_flags |= LOOKUP_REVAL; | |
398 | goto retry; | |
399 | } | |
6902d925 | 400 | out: |
d84f4f99 DH |
401 | revert_creds(old_cred); |
402 | put_cred(override_cred); | |
1da177e4 LT |
403 | return res; |
404 | } | |
405 | ||
ca013e94 | 406 | SYSCALL_DEFINE2(access, const char __user *, filename, int, mode) |
5590ff0d UD |
407 | { |
408 | return sys_faccessat(AT_FDCWD, filename, mode); | |
409 | } | |
410 | ||
3cdad428 | 411 | SYSCALL_DEFINE1(chdir, const char __user *, filename) |
1da177e4 | 412 | { |
2d8f3038 | 413 | struct path path; |
1da177e4 | 414 | int error; |
0291c0a5 JL |
415 | unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY; |
416 | retry: | |
417 | error = user_path_at(AT_FDCWD, filename, lookup_flags, &path); | |
1da177e4 LT |
418 | if (error) |
419 | goto out; | |
420 | ||
9cfcac81 | 421 | error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR); |
1da177e4 LT |
422 | if (error) |
423 | goto dput_and_out; | |
424 | ||
2d8f3038 | 425 | set_fs_pwd(current->fs, &path); |
1da177e4 LT |
426 | |
427 | dput_and_out: | |
2d8f3038 | 428 | path_put(&path); |
0291c0a5 JL |
429 | if (retry_estale(error, lookup_flags)) { |
430 | lookup_flags |= LOOKUP_REVAL; | |
431 | goto retry; | |
432 | } | |
1da177e4 LT |
433 | out: |
434 | return error; | |
435 | } | |
436 | ||
3cdad428 | 437 | SYSCALL_DEFINE1(fchdir, unsigned int, fd) |
1da177e4 | 438 | { |
2903ff01 | 439 | struct fd f = fdget_raw(fd); |
1da177e4 | 440 | struct inode *inode; |
2903ff01 | 441 | int error = -EBADF; |
1da177e4 LT |
442 | |
443 | error = -EBADF; | |
2903ff01 | 444 | if (!f.file) |
1da177e4 LT |
445 | goto out; |
446 | ||
496ad9aa | 447 | inode = file_inode(f.file); |
1da177e4 LT |
448 | |
449 | error = -ENOTDIR; | |
450 | if (!S_ISDIR(inode->i_mode)) | |
451 | goto out_putf; | |
452 | ||
9cfcac81 | 453 | error = inode_permission(inode, MAY_EXEC | MAY_CHDIR); |
1da177e4 | 454 | if (!error) |
2903ff01 | 455 | set_fs_pwd(current->fs, &f.file->f_path); |
1da177e4 | 456 | out_putf: |
2903ff01 | 457 | fdput(f); |
1da177e4 LT |
458 | out: |
459 | return error; | |
460 | } | |
461 | ||
3480b257 | 462 | SYSCALL_DEFINE1(chroot, const char __user *, filename) |
1da177e4 | 463 | { |
2d8f3038 | 464 | struct path path; |
1da177e4 | 465 | int error; |
2771261e JL |
466 | unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY; |
467 | retry: | |
468 | error = user_path_at(AT_FDCWD, filename, lookup_flags, &path); | |
1da177e4 LT |
469 | if (error) |
470 | goto out; | |
471 | ||
9cfcac81 | 472 | error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR); |
1da177e4 LT |
473 | if (error) |
474 | goto dput_and_out; | |
475 | ||
476 | error = -EPERM; | |
c7b96acf | 477 | if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT)) |
1da177e4 | 478 | goto dput_and_out; |
8b8efb44 TH |
479 | error = security_path_chroot(&path); |
480 | if (error) | |
481 | goto dput_and_out; | |
1da177e4 | 482 | |
2d8f3038 | 483 | set_fs_root(current->fs, &path); |
1da177e4 LT |
484 | error = 0; |
485 | dput_and_out: | |
2d8f3038 | 486 | path_put(&path); |
2771261e JL |
487 | if (retry_estale(error, lookup_flags)) { |
488 | lookup_flags |= LOOKUP_REVAL; | |
489 | goto retry; | |
490 | } | |
1da177e4 LT |
491 | out: |
492 | return error; | |
493 | } | |
494 | ||
e57712eb | 495 | static int chmod_common(struct path *path, umode_t mode) |
1da177e4 | 496 | { |
e57712eb | 497 | struct inode *inode = path->dentry->d_inode; |
27ac0ffe | 498 | struct inode *delegated_inode = NULL; |
1da177e4 | 499 | struct iattr newattrs; |
e57712eb | 500 | int error; |
1da177e4 | 501 | |
e57712eb AV |
502 | error = mnt_want_write(path->mnt); |
503 | if (error) | |
504 | return error; | |
27ac0ffe | 505 | retry_deleg: |
fe542cf5 | 506 | mutex_lock(&inode->i_mutex); |
cdcf116d | 507 | error = security_path_chmod(path, mode); |
e57712eb | 508 | if (error) |
fe542cf5 | 509 | goto out_unlock; |
1da177e4 LT |
510 | newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO); |
511 | newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; | |
27ac0ffe | 512 | error = notify_change(path->dentry, &newattrs, &delegated_inode); |
fe542cf5 | 513 | out_unlock: |
1b1dcc1b | 514 | mutex_unlock(&inode->i_mutex); |
27ac0ffe BF |
515 | if (delegated_inode) { |
516 | error = break_deleg_wait(&delegated_inode); | |
517 | if (!error) | |
518 | goto retry_deleg; | |
519 | } | |
e57712eb AV |
520 | mnt_drop_write(path->mnt); |
521 | return error; | |
522 | } | |
523 | ||
49f0a076 | 524 | SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode) |
e57712eb | 525 | { |
173c8401 | 526 | struct fd f = fdget(fd); |
e57712eb AV |
527 | int err = -EBADF; |
528 | ||
173c8401 | 529 | if (f.file) { |
9f45f5bf | 530 | audit_file(f.file); |
173c8401 AV |
531 | err = chmod_common(&f.file->f_path, mode); |
532 | fdput(f); | |
e57712eb | 533 | } |
1da177e4 LT |
534 | return err; |
535 | } | |
536 | ||
49f0a076 | 537 | SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode) |
1da177e4 | 538 | { |
2d8f3038 | 539 | struct path path; |
1da177e4 | 540 | int error; |
14ff690c JL |
541 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
542 | retry: | |
543 | error = user_path_at(dfd, filename, lookup_flags, &path); | |
e57712eb AV |
544 | if (!error) { |
545 | error = chmod_common(&path, mode); | |
546 | path_put(&path); | |
14ff690c JL |
547 | if (retry_estale(error, lookup_flags)) { |
548 | lookup_flags |= LOOKUP_REVAL; | |
549 | goto retry; | |
550 | } | |
e57712eb | 551 | } |
1da177e4 LT |
552 | return error; |
553 | } | |
554 | ||
49f0a076 | 555 | SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode) |
5590ff0d UD |
556 | { |
557 | return sys_fchmodat(AT_FDCWD, filename, mode); | |
558 | } | |
559 | ||
fe542cf5 | 560 | static int chown_common(struct path *path, uid_t user, gid_t group) |
1da177e4 | 561 | { |
fe542cf5 | 562 | struct inode *inode = path->dentry->d_inode; |
27ac0ffe | 563 | struct inode *delegated_inode = NULL; |
1da177e4 LT |
564 | int error; |
565 | struct iattr newattrs; | |
52137abe EB |
566 | kuid_t uid; |
567 | kgid_t gid; | |
568 | ||
569 | uid = make_kuid(current_user_ns(), user); | |
570 | gid = make_kgid(current_user_ns(), group); | |
1da177e4 | 571 | |
1da177e4 LT |
572 | newattrs.ia_valid = ATTR_CTIME; |
573 | if (user != (uid_t) -1) { | |
52137abe EB |
574 | if (!uid_valid(uid)) |
575 | return -EINVAL; | |
1da177e4 | 576 | newattrs.ia_valid |= ATTR_UID; |
52137abe | 577 | newattrs.ia_uid = uid; |
1da177e4 LT |
578 | } |
579 | if (group != (gid_t) -1) { | |
52137abe EB |
580 | if (!gid_valid(gid)) |
581 | return -EINVAL; | |
1da177e4 | 582 | newattrs.ia_valid |= ATTR_GID; |
52137abe | 583 | newattrs.ia_gid = gid; |
1da177e4 LT |
584 | } |
585 | if (!S_ISDIR(inode->i_mode)) | |
b5376771 SH |
586 | newattrs.ia_valid |= |
587 | ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV; | |
27ac0ffe | 588 | retry_deleg: |
1b1dcc1b | 589 | mutex_lock(&inode->i_mutex); |
d2b31ca6 | 590 | error = security_path_chown(path, uid, gid); |
fe542cf5 | 591 | if (!error) |
27ac0ffe | 592 | error = notify_change(path->dentry, &newattrs, &delegated_inode); |
1b1dcc1b | 593 | mutex_unlock(&inode->i_mutex); |
27ac0ffe BF |
594 | if (delegated_inode) { |
595 | error = break_deleg_wait(&delegated_inode); | |
596 | if (!error) | |
597 | goto retry_deleg; | |
598 | } | |
1da177e4 LT |
599 | return error; |
600 | } | |
601 | ||
6559eed8 HC |
602 | SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user, |
603 | gid_t, group, int, flag) | |
5590ff0d | 604 | { |
2d8f3038 | 605 | struct path path; |
5590ff0d | 606 | int error = -EINVAL; |
65cfc672 | 607 | int lookup_flags; |
5590ff0d | 608 | |
65cfc672 | 609 | if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0) |
5590ff0d UD |
610 | goto out; |
611 | ||
65cfc672 AV |
612 | lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW; |
613 | if (flag & AT_EMPTY_PATH) | |
614 | lookup_flags |= LOOKUP_EMPTY; | |
99a5df37 | 615 | retry: |
65cfc672 | 616 | error = user_path_at(dfd, filename, lookup_flags, &path); |
6902d925 DH |
617 | if (error) |
618 | goto out; | |
2d8f3038 | 619 | error = mnt_want_write(path.mnt); |
2af482a7 DH |
620 | if (error) |
621 | goto out_release; | |
fe542cf5 | 622 | error = chown_common(&path, user, group); |
2d8f3038 | 623 | mnt_drop_write(path.mnt); |
2af482a7 | 624 | out_release: |
2d8f3038 | 625 | path_put(&path); |
99a5df37 JL |
626 | if (retry_estale(error, lookup_flags)) { |
627 | lookup_flags |= LOOKUP_REVAL; | |
628 | goto retry; | |
629 | } | |
5590ff0d UD |
630 | out: |
631 | return error; | |
632 | } | |
633 | ||
55e4def0 | 634 | SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group) |
1da177e4 | 635 | { |
55e4def0 DH |
636 | return sys_fchownat(AT_FDCWD, filename, user, group, 0); |
637 | } | |
1da177e4 | 638 | |
55e4def0 DH |
639 | SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group) |
640 | { | |
641 | return sys_fchownat(AT_FDCWD, filename, user, group, | |
642 | AT_SYMLINK_NOFOLLOW); | |
1da177e4 LT |
643 | } |
644 | ||
ca013e94 | 645 | SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group) |
1da177e4 | 646 | { |
2903ff01 | 647 | struct fd f = fdget(fd); |
1da177e4 LT |
648 | int error = -EBADF; |
649 | ||
2903ff01 | 650 | if (!f.file) |
6902d925 DH |
651 | goto out; |
652 | ||
2903ff01 | 653 | error = mnt_want_write_file(f.file); |
2af482a7 DH |
654 | if (error) |
655 | goto out_fput; | |
9f45f5bf | 656 | audit_file(f.file); |
2903ff01 AV |
657 | error = chown_common(&f.file->f_path, user, group); |
658 | mnt_drop_write_file(f.file); | |
2af482a7 | 659 | out_fput: |
2903ff01 | 660 | fdput(f); |
6902d925 | 661 | out: |
1da177e4 LT |
662 | return error; |
663 | } | |
664 | ||
90ad1a8e MS |
665 | int open_check_o_direct(struct file *f) |
666 | { | |
667 | /* NB: we're sure to have correct a_ops only after f_op->open */ | |
668 | if (f->f_flags & O_DIRECT) { | |
669 | if (!f->f_mapping->a_ops || | |
670 | ((!f->f_mapping->a_ops->direct_IO) && | |
671 | (!f->f_mapping->a_ops->get_xip_mem))) { | |
672 | return -EINVAL; | |
673 | } | |
674 | } | |
675 | return 0; | |
676 | } | |
677 | ||
02e5180d | 678 | static int do_dentry_open(struct file *f, |
96b7e579 AV |
679 | int (*open)(struct inode *, struct file *), |
680 | const struct cred *cred) | |
1da177e4 | 681 | { |
1abf0c71 | 682 | static const struct file_operations empty_fops = {}; |
1da177e4 LT |
683 | struct inode *inode; |
684 | int error; | |
685 | ||
5300990c | 686 | f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK | |
a1a5b3d9 | 687 | FMODE_PREAD | FMODE_PWRITE; |
1abf0c71 | 688 | |
b5bcdda3 | 689 | path_get(&f->f_path); |
dd37978c | 690 | inode = f->f_inode = f->f_path.dentry->d_inode; |
1da177e4 | 691 | f->f_mapping = inode->i_mapping; |
1da177e4 | 692 | |
3f4d5a00 AV |
693 | if (unlikely(f->f_flags & O_PATH)) { |
694 | f->f_mode = FMODE_PATH; | |
1abf0c71 | 695 | f->f_op = &empty_fops; |
96b7e579 | 696 | return 0; |
1abf0c71 AV |
697 | } |
698 | ||
dd20908a | 699 | if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) { |
0ccb2863 | 700 | error = get_write_access(inode); |
3f4d5a00 | 701 | if (unlikely(error)) |
1da177e4 | 702 | goto cleanup_file; |
0ccb2863 | 703 | error = __mnt_want_write(f->f_path.mnt); |
3f4d5a00 | 704 | if (unlikely(error)) { |
0ccb2863 AV |
705 | put_write_access(inode); |
706 | goto cleanup_file; | |
707 | } | |
83f936c7 | 708 | f->f_mode |= FMODE_WRITER; |
1da177e4 LT |
709 | } |
710 | ||
9c225f26 LT |
711 | /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */ |
712 | if (S_ISREG(inode->i_mode)) | |
713 | f->f_mode |= FMODE_ATOMIC_POS; | |
714 | ||
1abf0c71 | 715 | f->f_op = fops_get(inode->i_fop); |
72c2d531 AV |
716 | if (unlikely(WARN_ON(!f->f_op))) { |
717 | error = -ENODEV; | |
718 | goto cleanup_all; | |
719 | } | |
1abf0c71 | 720 | |
83d49856 | 721 | error = security_file_open(f, cred); |
788e7dd4 YN |
722 | if (error) |
723 | goto cleanup_all; | |
724 | ||
f3c7691e BF |
725 | error = break_lease(inode, f->f_flags); |
726 | if (error) | |
727 | goto cleanup_all; | |
728 | ||
72c2d531 | 729 | if (!open) |
834f2a4a TM |
730 | open = f->f_op->open; |
731 | if (open) { | |
732 | error = open(inode, f); | |
1da177e4 LT |
733 | if (error) |
734 | goto cleanup_all; | |
735 | } | |
890275b5 MZ |
736 | if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ) |
737 | i_readcount_inc(inode); | |
293bc982 AV |
738 | if ((f->f_mode & FMODE_READ) && |
739 | likely(f->f_op->read || f->f_op->aio_read || f->f_op->read_iter)) | |
7f7f25e8 | 740 | f->f_mode |= FMODE_CAN_READ; |
293bc982 AV |
741 | if ((f->f_mode & FMODE_WRITE) && |
742 | likely(f->f_op->write || f->f_op->aio_write || f->f_op->write_iter)) | |
7f7f25e8 | 743 | f->f_mode |= FMODE_CAN_WRITE; |
834f2a4a | 744 | |
1da177e4 LT |
745 | f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC); |
746 | ||
747 | file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping); | |
748 | ||
96b7e579 | 749 | return 0; |
1da177e4 LT |
750 | |
751 | cleanup_all: | |
752 | fops_put(f->f_op); | |
83f936c7 | 753 | if (f->f_mode & FMODE_WRITER) { |
1da177e4 | 754 | put_write_access(inode); |
83f936c7 | 755 | __mnt_drop_write(f->f_path.mnt); |
4a3fd211 | 756 | } |
1da177e4 | 757 | cleanup_file: |
02e5180d AV |
758 | path_put(&f->f_path); |
759 | f->f_path.mnt = NULL; | |
760 | f->f_path.dentry = NULL; | |
dd37978c | 761 | f->f_inode = NULL; |
96b7e579 | 762 | return error; |
1da177e4 LT |
763 | } |
764 | ||
d18e9008 MS |
765 | /** |
766 | * finish_open - finish opening a file | |
0854d450 | 767 | * @file: file pointer |
d18e9008 MS |
768 | * @dentry: pointer to dentry |
769 | * @open: open callback | |
0854d450 | 770 | * @opened: state of open |
d18e9008 MS |
771 | * |
772 | * This can be used to finish opening a file passed to i_op->atomic_open(). | |
773 | * | |
774 | * If the open callback is set to NULL, then the standard f_op->open() | |
775 | * filesystem callback is substituted. | |
0854d450 MS |
776 | * |
777 | * NB: the dentry reference is _not_ consumed. If, for example, the dentry is | |
778 | * the return value of d_splice_alias(), then the caller needs to perform dput() | |
779 | * on it after finish_open(). | |
780 | * | |
781 | * On successful return @file is a fully instantiated open file. After this, if | |
782 | * an error occurs in ->atomic_open(), it needs to clean up with fput(). | |
783 | * | |
784 | * Returns zero on success or -errno if the open failed. | |
d18e9008 | 785 | */ |
30d90494 AV |
786 | int finish_open(struct file *file, struct dentry *dentry, |
787 | int (*open)(struct inode *, struct file *), | |
788 | int *opened) | |
d18e9008 | 789 | { |
96b7e579 | 790 | int error; |
3d8a00d2 | 791 | BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */ |
d18e9008 | 792 | |
b5bcdda3 | 793 | file->f_path.dentry = dentry; |
02e5180d | 794 | error = do_dentry_open(file, open, current_cred()); |
96b7e579 | 795 | if (!error) |
47237687 | 796 | *opened |= FILE_OPENED; |
d18e9008 | 797 | |
96b7e579 | 798 | return error; |
d18e9008 MS |
799 | } |
800 | EXPORT_SYMBOL(finish_open); | |
801 | ||
802 | /** | |
803 | * finish_no_open - finish ->atomic_open() without opening the file | |
804 | * | |
0854d450 | 805 | * @file: file pointer |
d18e9008 MS |
806 | * @dentry: dentry or NULL (as returned from ->lookup()) |
807 | * | |
808 | * This can be used to set the result of a successful lookup in ->atomic_open(). | |
0854d450 MS |
809 | * |
810 | * NB: unlike finish_open() this function does consume the dentry reference and | |
811 | * the caller need not dput() it. | |
812 | * | |
813 | * Returns "1" which must be the return value of ->atomic_open() after having | |
814 | * called this function. | |
d18e9008 | 815 | */ |
e45198a6 | 816 | int finish_no_open(struct file *file, struct dentry *dentry) |
d18e9008 | 817 | { |
30d90494 | 818 | file->f_path.dentry = dentry; |
e45198a6 | 819 | return 1; |
d18e9008 MS |
820 | } |
821 | EXPORT_SYMBOL(finish_no_open); | |
822 | ||
765927b2 | 823 | struct file *dentry_open(const struct path *path, int flags, |
745ca247 | 824 | const struct cred *cred) |
a1a5b3d9 PS |
825 | { |
826 | int error; | |
827 | struct file *f; | |
828 | ||
e0e81739 DH |
829 | validate_creds(cred); |
830 | ||
c212f9aa | 831 | /* We must always pass in a valid mount pointer. */ |
765927b2 | 832 | BUG_ON(!path->mnt); |
322ee5b3 | 833 | |
a1a5b3d9 | 834 | f = get_empty_filp(); |
1afc99be AV |
835 | if (!IS_ERR(f)) { |
836 | f->f_flags = flags; | |
4aa7c634 | 837 | error = vfs_open(path, f, cred); |
1afc99be AV |
838 | if (!error) { |
839 | /* from now on we need fput() to dispose of f */ | |
840 | error = open_check_o_direct(f); | |
841 | if (error) { | |
842 | fput(f); | |
843 | f = ERR_PTR(error); | |
844 | } | |
845 | } else { | |
846 | put_filp(f); | |
2a027e7a AV |
847 | f = ERR_PTR(error); |
848 | } | |
2a027e7a AV |
849 | } |
850 | return f; | |
a1a5b3d9 | 851 | } |
1da177e4 LT |
852 | EXPORT_SYMBOL(dentry_open); |
853 | ||
4aa7c634 MS |
854 | /** |
855 | * vfs_open - open the file at the given path | |
856 | * @path: path to open | |
857 | * @filp: newly allocated file with f_flag initialized | |
858 | * @cred: credentials to use | |
859 | */ | |
860 | int vfs_open(const struct path *path, struct file *filp, | |
861 | const struct cred *cred) | |
862 | { | |
863 | struct inode *inode = path->dentry->d_inode; | |
864 | ||
865 | if (inode->i_op->dentry_open) | |
866 | return inode->i_op->dentry_open(path->dentry, filp, cred); | |
867 | else { | |
868 | filp->f_path = *path; | |
869 | return do_dentry_open(filp, NULL, cred); | |
870 | } | |
871 | } | |
872 | EXPORT_SYMBOL(vfs_open); | |
873 | ||
a218d0fd | 874 | static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op) |
47c805dc AV |
875 | { |
876 | int lookup_flags = 0; | |
877 | int acc_mode; | |
878 | ||
e305f48b | 879 | if (flags & (O_CREAT | __O_TMPFILE)) |
e68726ff MS |
880 | op->mode = (mode & S_IALLUGO) | S_IFREG; |
881 | else | |
882 | op->mode = 0; | |
47c805dc AV |
883 | |
884 | /* Must never be set by userspace */ | |
c6f3d811 | 885 | flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC; |
47c805dc AV |
886 | |
887 | /* | |
888 | * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only | |
889 | * check for O_DSYNC if the need any syncing at all we enforce it's | |
890 | * always set instead of having to deal with possibly weird behaviour | |
891 | * for malicious applications setting only __O_SYNC. | |
892 | */ | |
893 | if (flags & __O_SYNC) | |
894 | flags |= O_DSYNC; | |
895 | ||
bb458c64 AV |
896 | if (flags & __O_TMPFILE) { |
897 | if ((flags & O_TMPFILE_MASK) != O_TMPFILE) | |
60545d0d AV |
898 | return -EINVAL; |
899 | acc_mode = MAY_OPEN | ACC_MODE(flags); | |
ba57ea64 AV |
900 | if (!(acc_mode & MAY_WRITE)) |
901 | return -EINVAL; | |
60545d0d AV |
902 | } else if (flags & O_PATH) { |
903 | /* | |
904 | * If we have O_PATH in the open flag. Then we | |
905 | * cannot have anything other than the below set of flags | |
906 | */ | |
1abf0c71 AV |
907 | flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH; |
908 | acc_mode = 0; | |
909 | } else { | |
910 | acc_mode = MAY_OPEN | ACC_MODE(flags); | |
911 | } | |
47c805dc | 912 | |
1abf0c71 | 913 | op->open_flag = flags; |
47c805dc AV |
914 | |
915 | /* O_TRUNC implies we need access checks for write permissions */ | |
916 | if (flags & O_TRUNC) | |
917 | acc_mode |= MAY_WRITE; | |
918 | ||
919 | /* Allow the LSM permission hook to distinguish append | |
920 | access from general write access. */ | |
921 | if (flags & O_APPEND) | |
922 | acc_mode |= MAY_APPEND; | |
923 | ||
924 | op->acc_mode = acc_mode; | |
925 | ||
1abf0c71 AV |
926 | op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN; |
927 | ||
47c805dc AV |
928 | if (flags & O_CREAT) { |
929 | op->intent |= LOOKUP_CREATE; | |
930 | if (flags & O_EXCL) | |
931 | op->intent |= LOOKUP_EXCL; | |
932 | } | |
933 | ||
934 | if (flags & O_DIRECTORY) | |
935 | lookup_flags |= LOOKUP_DIRECTORY; | |
936 | if (!(flags & O_NOFOLLOW)) | |
937 | lookup_flags |= LOOKUP_FOLLOW; | |
f9652e10 AV |
938 | op->lookup_flags = lookup_flags; |
939 | return 0; | |
47c805dc AV |
940 | } |
941 | ||
669abf4e JL |
942 | /** |
943 | * file_open_name - open file and return file pointer | |
944 | * | |
945 | * @name: struct filename containing path to open | |
946 | * @flags: open flags as per the open(2) second argument | |
947 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
948 | * | |
949 | * This is the helper to open a file from kernelspace if you really | |
950 | * have to. But in generally you should not do this, so please move | |
951 | * along, nothing to see here.. | |
952 | */ | |
953 | struct file *file_open_name(struct filename *name, int flags, umode_t mode) | |
954 | { | |
955 | struct open_flags op; | |
f9652e10 AV |
956 | int err = build_open_flags(flags, mode, &op); |
957 | return err ? ERR_PTR(err) : do_filp_open(AT_FDCWD, name, &op); | |
669abf4e JL |
958 | } |
959 | ||
47c805dc AV |
960 | /** |
961 | * filp_open - open file and return file pointer | |
962 | * | |
963 | * @filename: path to open | |
964 | * @flags: open flags as per the open(2) second argument | |
965 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
966 | * | |
967 | * This is the helper to open a file from kernelspace if you really | |
968 | * have to. But in generally you should not do this, so please move | |
969 | * along, nothing to see here.. | |
970 | */ | |
a218d0fd | 971 | struct file *filp_open(const char *filename, int flags, umode_t mode) |
47c805dc | 972 | { |
669abf4e JL |
973 | struct filename name = {.name = filename}; |
974 | return file_open_name(&name, flags, mode); | |
47c805dc AV |
975 | } |
976 | EXPORT_SYMBOL(filp_open); | |
977 | ||
73d049a4 AV |
978 | struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt, |
979 | const char *filename, int flags) | |
980 | { | |
981 | struct open_flags op; | |
f9652e10 AV |
982 | int err = build_open_flags(flags, 0, &op); |
983 | if (err) | |
984 | return ERR_PTR(err); | |
73d049a4 AV |
985 | if (flags & O_CREAT) |
986 | return ERR_PTR(-EINVAL); | |
987 | if (!filename && (flags & O_DIRECTORY)) | |
988 | if (!dentry->d_inode->i_op->lookup) | |
989 | return ERR_PTR(-ENOTDIR); | |
f9652e10 | 990 | return do_file_open_root(dentry, mnt, filename, &op); |
73d049a4 AV |
991 | } |
992 | EXPORT_SYMBOL(file_open_root); | |
993 | ||
a218d0fd | 994 | long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode) |
1da177e4 | 995 | { |
47c805dc | 996 | struct open_flags op; |
f9652e10 AV |
997 | int fd = build_open_flags(flags, mode, &op); |
998 | struct filename *tmp; | |
999 | ||
1000 | if (fd) | |
1001 | return fd; | |
1002 | ||
1003 | tmp = getname(filename); | |
1004 | if (IS_ERR(tmp)) | |
1005 | return PTR_ERR(tmp); | |
1006 | ||
1007 | fd = get_unused_fd_flags(flags); | |
1008 | if (fd >= 0) { | |
1009 | struct file *f = do_filp_open(dfd, tmp, &op); | |
1010 | if (IS_ERR(f)) { | |
1011 | put_unused_fd(fd); | |
1012 | fd = PTR_ERR(f); | |
1013 | } else { | |
1014 | fsnotify_open(f); | |
1015 | fd_install(fd, f); | |
1da177e4 | 1016 | } |
1da177e4 | 1017 | } |
f9652e10 | 1018 | putname(tmp); |
1da177e4 | 1019 | return fd; |
1da177e4 | 1020 | } |
e922efc3 | 1021 | |
a218d0fd | 1022 | SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode) |
e922efc3 MS |
1023 | { |
1024 | if (force_o_largefile()) | |
1025 | flags |= O_LARGEFILE; | |
1026 | ||
2cf09666 | 1027 | return do_sys_open(AT_FDCWD, filename, flags, mode); |
e922efc3 | 1028 | } |
1da177e4 | 1029 | |
6559eed8 | 1030 | SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, |
a218d0fd | 1031 | umode_t, mode) |
5590ff0d UD |
1032 | { |
1033 | if (force_o_largefile()) | |
1034 | flags |= O_LARGEFILE; | |
1035 | ||
2cf09666 | 1036 | return do_sys_open(dfd, filename, flags, mode); |
5590ff0d | 1037 | } |
5590ff0d | 1038 | |
1da177e4 LT |
1039 | #ifndef __alpha__ |
1040 | ||
1041 | /* | |
1042 | * For backward compatibility? Maybe this should be moved | |
1043 | * into arch/i386 instead? | |
1044 | */ | |
a218d0fd | 1045 | SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode) |
1da177e4 LT |
1046 | { |
1047 | return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode); | |
1048 | } | |
1049 | ||
1050 | #endif | |
1051 | ||
1052 | /* | |
1053 | * "id" is the POSIX thread ID. We use the | |
1054 | * files pointer for this.. | |
1055 | */ | |
1056 | int filp_close(struct file *filp, fl_owner_t id) | |
1057 | { | |
45778ca8 | 1058 | int retval = 0; |
1da177e4 LT |
1059 | |
1060 | if (!file_count(filp)) { | |
1061 | printk(KERN_ERR "VFS: Close: file count is 0\n"); | |
45778ca8 | 1062 | return 0; |
1da177e4 LT |
1063 | } |
1064 | ||
72c2d531 | 1065 | if (filp->f_op->flush) |
75e1fcc0 | 1066 | retval = filp->f_op->flush(filp, id); |
1da177e4 | 1067 | |
1abf0c71 AV |
1068 | if (likely(!(filp->f_mode & FMODE_PATH))) { |
1069 | dnotify_flush(filp, id); | |
1070 | locks_remove_posix(filp, id); | |
1071 | } | |
1da177e4 LT |
1072 | fput(filp); |
1073 | return retval; | |
1074 | } | |
1075 | ||
1076 | EXPORT_SYMBOL(filp_close); | |
1077 | ||
1078 | /* | |
1079 | * Careful here! We test whether the file pointer is NULL before | |
1080 | * releasing the fd. This ensures that one clone task can't release | |
1081 | * an fd while another clone is opening it. | |
1082 | */ | |
ca013e94 | 1083 | SYSCALL_DEFINE1(close, unsigned int, fd) |
1da177e4 | 1084 | { |
483ce1d4 | 1085 | int retval = __close_fd(current->files, fd); |
ee731f4f EP |
1086 | |
1087 | /* can't restart close syscall because file table entry was cleared */ | |
1088 | if (unlikely(retval == -ERESTARTSYS || | |
1089 | retval == -ERESTARTNOINTR || | |
1090 | retval == -ERESTARTNOHAND || | |
1091 | retval == -ERESTART_RESTARTBLOCK)) | |
1092 | retval = -EINTR; | |
1093 | ||
1094 | return retval; | |
1da177e4 | 1095 | } |
1da177e4 LT |
1096 | EXPORT_SYMBOL(sys_close); |
1097 | ||
1098 | /* | |
1099 | * This routine simulates a hangup on the tty, to arrange that users | |
1100 | * are given clean terminals at login time. | |
1101 | */ | |
ca013e94 | 1102 | SYSCALL_DEFINE0(vhangup) |
1da177e4 LT |
1103 | { |
1104 | if (capable(CAP_SYS_TTY_CONFIG)) { | |
2cb5998b | 1105 | tty_vhangup_self(); |
1da177e4 LT |
1106 | return 0; |
1107 | } | |
1108 | return -EPERM; | |
1109 | } | |
1110 | ||
1111 | /* | |
1112 | * Called when an inode is about to be open. | |
1113 | * We use this to disallow opening large files on 32bit systems if | |
1114 | * the caller didn't specify O_LARGEFILE. On 64bit systems we force | |
1115 | * on this flag in sys_open. | |
1116 | */ | |
1117 | int generic_file_open(struct inode * inode, struct file * filp) | |
1118 | { | |
1119 | if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS) | |
a9c62a18 | 1120 | return -EOVERFLOW; |
1da177e4 LT |
1121 | return 0; |
1122 | } | |
1123 | ||
1124 | EXPORT_SYMBOL(generic_file_open); | |
1125 | ||
1126 | /* | |
1127 | * This is used by subsystems that don't want seekable | |
06b1e104 DT |
1128 | * file descriptors. The function is not supposed to ever fail, the only |
1129 | * reason it returns an 'int' and not 'void' is so that it can be plugged | |
1130 | * directly into file_operations structure. | |
1da177e4 LT |
1131 | */ |
1132 | int nonseekable_open(struct inode *inode, struct file *filp) | |
1133 | { | |
1134 | filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE); | |
1135 | return 0; | |
1136 | } | |
1137 | ||
1138 | EXPORT_SYMBOL(nonseekable_open); |